CSO (Composite Second-Order)
How Second-Order Beats Accumulate in a CATV Plant
When many RF carriers share a single broadband amplifier or directly modulate one optical transmitter, the device's small second-order nonlinearity mixes every pair of carriers together. Each mix produces sum and difference products at f1 + f2 and f1 − f2. In a real cable plant carrying 100-plus channels, thousands of these second-order products are generated, and many of them land inside the occupied bandwidth of other channels. CSO is the composite, or summed, power of all those second-order beats that fall close enough to a victim carrier to degrade it, referenced to that carrier's level and reported in dBc.
Because of the way standard channel plans are arranged, second-order products do not spread uniformly. In NTSC channelization the beats cluster at predictable offsets of roughly 0.75 MHz and 1.25 MHz from the visual carrier, producing two discrete CSO clusters near the band edges. This is the key visual signature on a spectrum analyzer: CSO appears as paired spikes offset from each carrier, whereas third-order composite triple beat fills the channel center. CSO tends to be worst at the extreme low and high ends of a wide downstream band, which is why operators sweep-test the full 54 MHz to 1002 MHz (or 1218 MHz) range with all carriers loaded.
The defining behavior of CSO is its dependence on drive level. Because second-order distortion power rises as the square of the fundamental amplitude, CSO worsens 2 dB for every 1 dB increase in per-channel operating level. This 2:1 slope distinguishes it from third-order products, which follow a 3:1 slope. Setting the amplifier operating point therefore becomes a balance: lower drive improves CSO and CTB but worsens the carrier-to-noise ratio, while higher drive improves carrier-to-noise but pushes distortion up.
Governing Equations
CSO = 10 log10(P2nd order beat / Pcarrier) dBc
Level dependence (per-channel drive change ΔL):
ΔCSO ≈ 2 × ΔL dB (vs. 3 × ΔL for CTB)
Cascade of N identical amplifier stages:
CSOcascade = CSOstage + k log10(N) dB (k = 20 coherent worst case, ~15 typical for CSO)
Where P values are RF powers, ΔL is the change in per-carrier RF level in dB, and N is the number of cascaded gain blocks. The coefficient k bounds how distortion adds down the cascade: k = 20 assumes the beats add in voltage (in phase), k = 10 assumes random-phase power addition. CTB is normally budgeted at the coherent k = 20; CSO beats are less perfectly aligned, so designers commonly use k ≈ 15. Worst-case example: a single stage at −65 dBc CSO cascaded 4 deep gives −65 + 20 log(4) ≈ −65 + 12 = −53 dBc end-of-line.
CSO vs. Other Distortion Metrics
| Metric | Beat Type | Slope vs. Level | Spectral Location | Typical Full-Load Spec |
|---|---|---|---|---|
| CSO | Second-order (f1 ± f2) | 2 dB / dB | ~0.75 & 1.25 MHz from carrier | −53 to −65 dBc |
| CTB | Third-order (f1 ± f2 ± f3) | 3 dB / dB | Near channel center | −53 to −65 dBc |
| IMD2 (two-tone) | Second-order | 2 dB / dB | Sum/difference of two tones | Component-level only |
| IMD3 (two-tone) | Third-order | 3 dB / dB | 2f1−f2, 2f2−f1 | Component-level only |
| XMOD | Cross-modulation | 2 dB / dB | Sync-rate sidebands | −57 dBc (analog) |
Frequently Asked Questions
How does CSO differ from CTB in a cable plant?
CSO sums the second-order products at f1 ± f2, which cluster about 0.75 and 1.25 MHz from each visual carrier and follow a 2 dB-per-dB level slope. CTB sums the third-order products at f1 ± f2 ± f3, which pile up near channel center and follow a 3 dB-per-dB slope. CSO usually dominates at the band edges of a wide downstream; CTB dominates in the loaded middle.
What CSO level is required for a 1 GHz HFC downstream?
Analog NTSC carriage calls for better than −53 dBc at the tap, with designers budgeting end-of-line CSO around −57 to −60 dBc for margin. All-QAM plants still spec each gain stage at −65 dBc or better at full load so an accumulated 4 to 6 amplifier cascade stays below the noise-equivalent floor. Cascading N identical stages degrades CSO by k log(N) dB, where k is 20 in the coherent worst case and about 15 in typical CSO practice.
Why does CSO get worse as you add more channels?
The number of second-order beats landing on a channel grows roughly linearly with loaded carriers, and every added carrier raises the composite RMS power driving the amplifier toward compression. Doubling the channel count adds about 3 dB of composite level, so measured CSO can worsen 5 to 8 dB. That is why CATV amplifiers are specified, and swept, at full channel loading.